A First Runtime Analysis of NSGA-III on a Many-Objective Multimodal Problem: Provable Exponential Speedup via Stochastic Population Update

2025-05-02

AI Summary

This work addresses the lack of rigorous runtime analysis for NSGA-III on multi-objective multimodal optimization problems—particularly the OneJumpZeroJump (OjZj) benchmark—where theoretical guarantees on convergence and efficiency remain absent. Method: We conduct the first tight runtime analysis by integrating population dynamics modeling, quantification of selection pressure, and discrete probabilistic analysis. Contribution/Results: We derive a tight upper bound of $O(n^{k+d/2} + mu n ln n)$ on the expected runtime for constant numbers of objectives. Theoretically, when the population size satisfies $mu = omega(n^{d/2})$, NSGA-III achieves exponential speedup over NSGA-II. Furthermore, we propose and rigorously verify a stochastic population update strategy that yields a provable acceleration factor of $Theta((k/b)^{k-1})$. This study fills a fundamental theoretical gap in understanding NSGA-III’s convergence mechanism and overcomes a key bottleneck in the runtime analysis of multi-objective evolutionary algorithms.

Technology Category

• Search and Optimization: Mixed Discrete/Continuous Search
• Reasoning under Uncertainty: Stochastic Optimization
• Constraint Satisfaction and Optimization: Other Foundations of Constraint Satisfaction

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Web Mining and Content Analysis: Models for Web evolution

Abstract

The NSGA-III is a prominent algorithm in evolutionary many-objective optimization. It is well-suited for optimizing functions with more than three objectives, setting it apart from the classic NSGA-II. However, theoretical insights about NSGA-III of when and why it performs well are still in its early development. This paper addresses this point and conducts a rigorous runtime analysis of NSGA-III on the many-objective extsc{OneJumpZeroJump} benchmark ( extsc{OjZj} for short), providing the first runtime bounds where the number of objectives is constant. We show that NSGA-III finds the Pareto front of extsc{OjZj} in time $O(n^{k+d/2}+ mu n ln(n))$ where $n$ is the problem size, $d$ is the number of objectives, $k$ is the gap size, a problem specific parameter, if its population size $mu in 2^{O(n)}$ is at least $(2n/d+1)^{d/2}$. Notably, NSGA-III is faster than NSGA-II by a factor of $mu/n^{d/2}$ for some $mu in omega(n^{d/2})$. We also show that a stochastic population update, proposed by Bian et al., provably guarantees a speedup of order $Theta((k/b)^{k-1})$ in the runtime where $b>0$ is a constant. To our knowledge, this is the first rigorous runtime analysis of NSGA-III on extsc{OjZj}. Proving these bounds requires a much deeper understanding of the population dynamics of NSGA-III than previous papers achieved.

Problem

• Analyzes NSGA-III performance on many-objective multimodal problems
• Compares NSGA-III speed with NSGA-II theoretically
• Proves stochastic population update accelerates NSGA-III runtime

Innovation

• NSGA-III optimizes many-objective multimodal problems
• Stochastic population update ensures exponential speedup
• Runtime analysis proves NSGA-III outperforms NSGA-II